2 min read

Machine learning and email security: The smarter way to avoid data leaks

Posted by Rick Goud on 9th January 2023

Machine learning and email security: The smarter way to avoid data leaks

Today, every organization handles huge quantities of sensitive data that needs protecting - but we also have busy work days, and digital security isn’t always top of the agenda. 

With over 70% of all data incidents reported to the ICO the result of human error, we regularly speak with IT leaders who are at a loss as to how to counter these most common challenges. When it comes to sending emails to the wrong person, accidentally cc’ing recipients in place of bcc’ing, or sending the wrong attachment, we beat the drum for smart technology. But that doesn’t mean introducing yet another collaboration platform to your tech-stack.

No. We keep things simple - but incredibly smart.

Zivver integrates with leading email clients (Gmail, Microsoft 365) to enhance basic security functionality. This means employees can apply advanced encryption to sensitive emails, send up to 5TB securely from their email client (say goodbye to third-party file transfer sites), recall emails without time limits, and much more.

Smart, and getting smarter, our solution is powered by machine learning, meaning it learns from user behaviour and recommends when users should consider encrypting emails and applying multi-factor authentication controls to protect data in transit and when at rest. 

In simple terms, Zivvers empowers users to avoid the seemingly small mistakes with the biggest consequences by identifying sensitive data in the body and attachments of emails - and, unlike other vendors on the market, our technology is smart enough to identify personal information, sensitive medical and financial information, and more, intervening before mistakes happen.

It does this through data classification, a key function for organizations seeking to prevent data leaks. However, our classification method is far smarter than traditional approaches. Most widely adopted processes require administrators to spend valuable time manually labelling information. Other systems automatically classify information into ‘wordlists’ on the basis of whether a word (e.g. "confidential") or a pattern (e.g. a social security number) appears.   

In both cases, data is frequently incorrectly identified as sensitive. In fact, our research shows that even the most advanced wordlists classify information correctly in only 10% to 20% of cases. The reason for this is simple. A single word can often be interpreted as sensitive regardless of the wider context of a sentence or email.

Zivver Smart Classification automates classification methods and triples the accuracy of security alerts by leveraging millions of data points. This is the next generation of data classification, using advanced machine learning to learn the intricate differences between sensitive and non-sensitive data. This fundamental innovation enables organizations to more effectively prevent data leaks and ensure compliance.

The result? Employees aren’t hassled by unnecessary notifications, and administrators don’t need to spend hours manually classifying data points. And, most importantly, organizations can work with confidence that their data is secure, depending on the sensitivity of the data included. 

To learn more about how Zivver can reduce data leaks at your organization, watch a demo or get in touch.

Rick Goud avatar

Rick Goud

CIO & Founder

Published: 9th January 2023

Subscribe to our newsletter
Share this

Enjoy this article? Share the knowledge

Stay informed with Zivver

Subscribe to get more email security tips straight to your inbox.